CN100397390C - Digital frequency response compensator and arbitrary response generator system - Google Patents

Abstract

A digital signal processing system (1) capable of compensating for frequency response variations (5) and generating a response characteristic (8) that complies with a provided specification (14). The system automatically generates digital filters (10) to provide this compensation with almost any form of channel frequency response information and with user defined specifications. The capability of this system to trade-off noise performance, pulse response, and frequency response flatness in order to provide an optimized response is demonstrated. The system also provides feedback to the user on the final response characteristics (15).

Description

Numerical frequency response compensator and any response generator system

The full content of the application of following submission is hereby incorporated by: sequence number is No.09/669,955, the applying date is the U. S. application on September 26th, 2000; Sequence number is No.09/988,120, the applying date is the U. S. application in November 16 calendar year 2001; Sequence number is No.09/988,420, the applying date is the U. S. application in November 16 calendar year 2001.

Background of invention

The present invention relates to have digital signal processing (DSP) system of numerical frequency response compensator and any (arbitrary) response generator.In general, the present invention relates to have analog input signal, analog electronic equipment (for example, attenuator, booster element and impact damper) and the system of analog to digital converter (ADC) of sequence that analog input signal converted to the numeral of this input signal.The present invention is fit to be designed to have above-mentioned parts so that obtain waveform being used to watch, analyze, test and identify purpose, and the instrument of other various purposes.More particularly, the present invention is fit to digital sampling oscillograph (DSO), the especially DSO of superelevation bandwidth and sampling rate and monopulse DSO (sometimes being called as real-time DSO).Enough enough the crossing of these DSO energy are taken a sample and fidelity comes the digitized voltage waveform to catch the waveform with single trigger event.

Traditionally, DSO be always the slip-stick artist be used for checking signal mainly watch instrument.Along with at a high speed, the complicated application of waveform in current communication and data storage industry, de-emphasized simply the watching of waveform, but the DSO that can also carry out wave form analysis expressed bigger demand.The demand that increases day by day to the DSO analysis ability needs signal fidelity (being the digitized wave forms of better quality) largely.In more requirements of making signal fidelity, the demand of DSO with higher bandwidth and sampling rate is continued not decline equally.Regrettably, high speed signal needs the DSO of high bandwidth, for superelevation bandwidth, high sampling rate, DSO will pay extra cost in real time.

The about every 2-2 of the sampling rate of digital oscilloscope ^1/2Year will double, and bandwidth doubled in almost per 4 years.The increase of bandwidth is not need cost just to occur.Usually, analog component is put to good use their limit.Sometimes, peaking circuit is used to further broadening bandwidth.This expansion to bigger bandwidth usually is that cost occurs with the signal fidelity, especially in impulse response (overshoot and damped oscillation) and frequency response flatness scope.This be since pulse peaking often some is uncontrolled (promptly be difficult to system of peaking and keep flat response simultaneously).In addition, because analog component is forced into their bandwidth limit, if therefore this bandwidth is exceeded, then frequency response usually sharply descends.Therefore, the high bandwidth oscillograph no longer has mild frequency response roll-off characteristic.

Although there is this situation, the original expectation of DSO user does not change.Even users still expect low noise-for bandwidth double at every turn noise bring up to 2 square root doubly-and they expect that DSO has definite roll-off characteristic.

The fact that makes this situation become further complicated is that the design of high speed DSO relates to extremely a large amount of balance and compromise.Three kinds of main conventional measure-noises, frequency response and time domain responses of signal fidelity-all antagonism each other.As mentioned previously, by the bigger bandwidth of oscillograph expansion the noise in the input signal is increased.Can increase overshoot and damped oscillation with any departing from all of first order pole or duopole frequency response characteristic.The limit that the bandwidth of hardware component is expanded to they only can make problem more serious.Make the frequency response smooth meeting that flattens that impulse response is worsened.Improve the bandwidth (this bandwidth is undesirable always) that impulse response typically means lowering apparatus.Because DSO is general instrument, therefore can selects balance modestly, but always have many users not to be met.The unique selection of leaving the user for is between several fixed-bandwidth scopes, and they are got in touch in a simple RC circuit.Even according to the bandwidth restricted pattern, this response also usually still can not respond in compliance with first order pole well, and the variation more than the 0.5dB can be arranged.

In the exploitation that the vertical market of DSO is used, wherein oscillograph simulation, for example, special communication or disc driver passage are necessary with consistent (compliance) of specific response.This ability that passage is simulated provides the ability of quick prototyping and analysis.

Summary of the invention

Therefore, according to the present invention, provide a kind of and can compensate the parts that increase the degeneration cause (being frequency response flatness and/or consistent with specific Expected Response characteristic) owing to bandwidth.

According to the present invention, provide a kind of balance that can make about noise, flatness and/or pusle response characteristics in addition, rather than depended on the tunable component of static instrument characteristic.Thereby this tunable component allows to optimize this instrument at given measurement.

According to the present invention, provide a kind of ability of using this tunable component to come the response characteristic of instrument is fed back to the user.

According to the present invention, provide a kind of and can be calibrated in addition to change the parts of channel response characteristic.

A preferred embodiment of the present invention provides a kind of signal processing system that can compensate the channel response characteristic of input waveform.This system comprises input specification (specification), wave filter builder and wave filter.Input specification is used for the regulation Filter Design, and comprises that regulation is used for obtaining the channel response characteristic of response characteristic of the passage of input waveform, and be used for frequency response that regulation wishes and with the user specification of the consistent degree of the frequency response of this hope.The wave filter builder is that wave filter generates coefficient and exports the final properties specification.Wave filter has the compensating filter generator, be used for generating the coefficient that responds corresponding to compensation according to the inverse (inverse) of channel response characteristic, and the response filter generator, be used for according to the coefficient of user specification generation corresponding to the combination of ideal response and noise reduction response.Wave filter filtration input waveform and output device are hopeful the overall response waveform of frequency response.This wave filter comprises: the filter coefficient high-speed cache that is used for the coefficient that the memory filter builder generated; Be used for filtering the compensating filter part of importing waveform according to the coefficient that is stored in the filter coefficient high-speed cache corresponding to the compensation response; And having the response filtration grade that is used to filter the compensation waveform of partly exporting from described compensating filter and the response filter part of noise reduction level, described compensating filter is partly exported the overall response waveform.Response filter partly uses the coefficient that is stored in the combination that responds corresponding to ideal response and noise reduction in the filter coefficient high-speed cache to carry out filtering.

In another aspect of the present invention, wave filter can be embodied as infinite impulse response (IIR) wave filter or finite impulse response (FIR) (FIR) wave filter.

Aspect another, can pre-determine the channel response characteristic of the present invention according to a reference signal with by the reference signal that passage obtains.

Of the present invention also aspect one, user specification can comprise that bandwidth, response are optimized, compensation unanimity and wave filter are realized type.Response optimization can be the flatness optimization that the impulse response optimization that utilizes Besselworth (Bessel irrigates and thinks) wave filter realization, the noise performance optimization of using the realization of Butterworth (Butterworth) wave filter or use Butterworth wave filter are realized.Wave filter realizes that type can be finite impulse response (FIR) (FIR) or infinite impulse response (IIR).

According to an alternative embodiment of the invention, provide a kind of Signal Processing Element that filters the digital waveform of input.This element comprises wave filter builder, infinite impulse response (IIR) wave filter, finite impulse response (FIR) (FIR) wave filter and outlet selector switch.The wave filter builder is used for generating filter coefficient according to channel frequence response and user's response characteristic.Determine the channel frequence response according to response input and correction input.Infinite impulse response (IIR) wave filter has IIR input that is used for this input digit waveform and the IIR coefficient input of linking the wave filter builder.Iir filter filters waveform according to the filter coefficient that is generated by the wave filter builder from this input digit waveform generation IIR.Finite impulse response (FIR) (FIR) wave filter has FIR input that is used for this input digit waveform and the FIR coefficient input of linking the wave filter builder.The FIR wave filter filters waveform according to the filter coefficient that is generated by the wave filter builder from this input digit waveform generation FIR.The outlet selector switch selects IIR to filter waveform or FIR filters waveform for output.

In this embodiment, the wave filter builder detects the variation of the sampling rate of following input digit waveform, and this variation can require filter coefficient to be changed or to regenerate.The wave filter builder is that FIR wave filter or iir filter generate filter coefficient according to the outlet selector switch.The wave filter builder has passage, compensation, reshaper and noise reduction output, is used for assessing filtering performance.

At this embodiment on the other hand, response input is a known input response, and described correction input is one by measurement input response that input channel obtained.User's response characteristic is used to generate the filter coefficient corresponding to any response part of this wave filter.User's response characteristic comprises bandwidth, response optimization, compensation unanimity and wave filter realization type.Response optimization can be the impulse response optimization that utilizes the Besselworth wave filter to realize, the noise performance optimization of utilizing the Butterworth wave filter to realize, or the flatness optimization that utilizes the Butterworth wave filter to realize.Wave filter realizes that type can be FIR or IIR.

In yet another embodiment of the present invention, provide a kind of method of filtering the digital waveform of input with the response characteristic of compensation acquisition channel.This method is at first according to the channel response of importing, generate the compensated part of a wave filter by the channel response of pre-warpage (pre-warp) input, design an analog filter, this analog filter is inferred the input channel response of simulating this pre-warpage by making initial filtering, and the coefficient that the initial filtering of iteration is inferred is to minimize mean square deviation, this analog filter that reverses, and use bilinear transformation to come the analog filter of this counter-rotating of digitizing to produce the compensated part of this wave filter.Afterwards, this method generates any response part of this wave filter according to the user specification of input.At last, this method is used the compensated part of this wave filter and is filtered the digital waveform of input with any response part of this wave filter, thereby produces the digital waveform of the filtration with Expected Response characteristic.

According to this embodiment, any response part of wave filter is made up of reshaper and denoiser.The coefficient that initial filtering is inferred carries out iteration, the compensation unanimity of stipulating in the user specification of mean square deviation less than input.

Aspect another of this embodiment, this wave filter can be embodied as infinite impulse response (IIR) wave filter or finite impulse response (FIR) (FIR) wave filter.

Also aspect of this embodiment, can pre-determine the channel frequence response characteristic according to a reference signal with by the reference signal that passage obtains.

Aspect another of this embodiment, user specification can comprise that bandwidth, response are optimized, compensation unanimity and wave filter are realized type.Response optimization can be the impulse response optimization that utilizes the Besselworth wave filter to realize, the noise performance optimization of utilizing the Butterworth wave filter to realize, or the flatness optimization that utilizes the Butterworth wave filter to realize.Wave filter realizes that type can be finite impulse response (FIR) (FIR) or infinite impulse response (IIR).

Other purpose of the present invention and advantage will be conspicuous in a measure, and will become clear in a measure from detailed description and accompanying drawing.

Brief description of drawings

In order to understand the present invention more completely, with reference to following description and accompanying drawing, in these accompanying drawings:

Fig. 1 represents according to frequency response compensator of the present invention and any response generator system;

Fig. 2 illustrates frequency response and the compensation and the response generation system of passage;

Fig. 3 be illustrated in one handle in the web according to frequency response compensator of the present invention and any response generator system;

Fig. 4 represents the inner structure of parts 33 shown in Figure 3;

The channel frequence response that Fig. 5 represents a time domain reference signal that gets access to and therefrom calculates;

Fig. 6 represents a pre-warpage channel frequence response;

Fig. 7 is illustrated in the filter response for initial filtering supposition in the response approximate shown in Figure 6;

Fig. 8 represents the pole and zero position for the initial filtering supposition of Fig. 7;

Fig. 9 represents to simulate the frequency response of the digital filter with variable consistent degree of response shown in Figure 6;

Figure 10 is the enlarged drawing of the frequency band among Fig. 9, wherein forces consistent;

Figure 11 represents to compensate the frequency response of the digital filter with variable consistent degree of response shown in Figure 6;

Figure 12 is the enlarged drawing of the frequency band among Figure 11, wherein forces consistent;

Figure 13 represents the compensation filter error of the wave filter among Figure 12;

Figure 14 represents the compensation filter error as the function of filtering progression;

Figure 15 represents to be used to specify the configuration user interfaces of wave filter;

Figure 16 represents to be used to specify the advanced level user interface of wave filter;

Figure 17 is the process flow diagram of the process of design Besselworth wave filter;

Figure 18 represents the frequency response of Besselworth wave filter;

Figure 19 represents to arrange according to the calibration of the DSO of being used for of the present invention; And

Figure 20 represents according to an example output performance specification of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED

Preferred embodiment according to system of the present invention is described with reference to the accompanying drawings.

The present invention is a kind of signal processing system with numerical frequency response compensator and any response generator.The present invention includes the filtering operation of between any downstream of ADC and digitized wave forms, carrying out by the Signal Processing Element in the signalling channel of DSO.Between reading duration and/or after reading, and before the further processing of demonstration or waveform, carry out filtering operation.DSO generally has and is used to handle the waveform that obtains so that the high-performance CPU (central processing unit) (CPU) of analyzing or showing.Can realize with the software on this CPU in the DSO according to digital filtering operation of the present invention.

The purpose of digital filtering operation is in order to change the frequency response of DSO.This wave filter is designed such that DSO overall system (comprising passage input, digitizing element and the present invention) has specific assigned frequency response by adjusting its filtering characteristic.In other words, digital filter not only filters frequency response, but also makes total system have the frequency response of regulation.Most wave filters are designed to the signal that is input to this wave filter is had specific effect simply, but specific total system response is not provided.

Fig. 2 illustrates the step of use according to the frequency response of compensation of the present invention and response generation system processing DSO passage.Input response curve Figure 17 represents the frequency response of typical input channel.Notice that this passage does not have the flat frequency response.Represent typical desired frequency response with desirable response curve Figure 20.But, perhaps the user wants to specify other frequency response.

The present invention utilizes the digital filter shown in 18 that response curve Figure 17 is transformed into overall response curve Figure 16.Notice that this response is not definite ideal response 20, but it is possible optimal response under the given channel hardware bandwidth.Digital filter 18 is made up of three inner filtering stages: compensating filter 19, ideal response wave filter 20 and by (cutoff) wave filter 21.

Fig. 2 also expresses each grade in these filtering stages to the influence of signal by being illustrated in system's overall response on each continuous filtering level.By simply on each grade additional frequency respond the influence of expressing each filtering stage.Recalling original system responses represents with channel frequence response 17,22.Channel response 22 is at first handled by compensated stage 23, produces the channel response 27 of compensation, and channel response 27 is smooth basically.Compensation filter level 23 is anti-(opposite)/(inverse) reciprocal of channel response 22.Notice that several curve maps (for example channel response 22) all have dash area (25) on the right side of curve map.This shadow representation is unknown in that local response by decay like this so that definite response.Afterwards, compensation response 28 produces output response 29 by ideal response filtering stage 30.Yet, notice that this output response still has the unknown content in the shadow region.By uncertain with eliminate this by level 31 processing of carrying out.By producing known overall response 32.

Former trial in this type invention part failure, this is owing to realizing that compensating filter partly has difficulties.The design of compensating filter is difficult, because it is based on the channel response of alterable height.

Although final wave filter be do as a whole realization and do not consider the effect of each grade, Filter Design will be divided into for two steps.This is because response filter is designed with cut-off filter.This two-step approach has been simplified Filter Design and has been reduced the filtering computing time of operating period.If this is because the channel frequence response changes, and has only compensating filter partly to need to rebuild so.Similarly, if the user changes response specifications, have only the response part to need to rebuild so.In other words, the design of compensation and response filter is separated.Output by compensating filter is designed to generate the fact of fixing output frequency response specifications (being flat response) and has proved this separation.Therefore, the design of this filter segment has been determined in the channel frequence response.Response filter supposes that partly its input response is smooth; Thereby only response output specification can influence design.

Fig. 1 represents according to frequency response compensator of the present invention and any response generator system 1.Notice that wave filter 4 is made up of previously discussed three filtering stages.Waveform is input to system and the above-mentioned filtering stage of process in input 2: compensation 5, response 6 and noise reduction sound 7 (or by).To respond 6 and noise reduction sound 7 combine filter segment 8 in response.The wave filter details table is shown infinite impulse response (IIR) dual quadrant part, and this is preferred, but does not require realization.

Wave filter 4 comprises filter coefficient high-speed cache 9, and high-speed cache 9 contains the coefficient that defines wave filter.Provide filter coefficient by filter coefficient builder (or wave filter builder) 10.Wave filter builder 10 separated into two parts: compensating filter generator 11 and response filter generator 12.Compensating filter generator 11 generates the filter coefficient of the compensation response that is used in Fig. 2.Response filter generator 12 generates the filter coefficient of the combination that is used for ideal response 20 and noise reduction sound (or ending) response 21, as shown in Figure 2.

The input specification of wave filter builder is made up of two parts: channel response characteristic 13 and response and compensation specification 14.Channel response specification 13 is based on the response of input channel, and response is to be stipulated by the user with compensation specification 14.The response that response and compensation specification 14 regulation are desired and with the consistent degree of the expectation of this response.Channel response 13 and user specification 14 have been stipulated the system performance of expectation fully.Channel response can be determined by the calibration of dispatching from the factory, or adopt normative reference to carry out dynamic calibration.Under the situation of dynamic calibration, this normative reference can be that inside provides, or this outside, unit.Response and compensation specification 14 are divided into grade to allow the tight control to Expected Response, allow easy control simultaneously this system.

Consistent degree allows the user fine to adjust this system.If the Expected Response specification is external,, will cause huge, complex digital filter so if perhaps Qi Wang consistent degree is very high.Such wave filter needs a large amount of processing times, the renewal rate of this meeting lowering apparatus.Therefore, consistent degree should balance each other with the influence to the instrument renewal rate.

Compensating filter generator 11 is set up the compensated part of wave filter.As previously discussed, this part is actually the inverse of channel response.Main difficulty on this partial design be channel response may some at random, and this specification can need to show the response of whole channel frequence.Minimum variance (squares error) between this final output that just causes Design of Filter to relate to input specification and wave filter responds (L2).Regrettably, when being indicated as L2 and minimizing, produced one group of nonlinear equation, they must use the nonlinear equation solution to solve.Equation be this non-linear fact mean do not guarantee L2 will be minimized-just will find the part to minimize.Though this can obtain handling under laboratory environment, thrashing is not allowed in the real world applications of majority.In addition, this instrument does not have the uncertain time amount of the calculation of filtered of being used for.Therefore, must take measures to maximize the chance of success of Design of Filter, and come calculating filter by immediate mode.

Response filter generator 12 translation (translate) user specifications 14 are also set up response filter.The combination of the normally general filter type with iir filter design (for example Butterworth, Bessel, Chebyshev reciprocal (Chebyshev)) compatibility of response filter.Also can use other filter type.

Following preferred embodiment is described and has been explained how the present invention handles this wave filter and set up problem.In case set up filter segment, wave filter just is cascaded (cascade) and can constantly filters the input waveform, and the total system response as appointment is provided.Owing to be included in the uncertainty in the compensating filter design partly, so the user should have the ability of checking final total systems performance.For this reason, provide one group of final specification 15 as the feedback of giving the user based on designed wave filter.

In being used for the new Software Development Platform of LeCroy DSO, realized the present invention.The principal character of software platform used in the present invention is that " streaming architecture " and " handling web "-the two all comprises the interconnection of a management processing object and the system that passes through the streams data of these objects.Referring to the applying date is that the sequence number in November 16 calendar year 2001 is No.09/988,120 U. S. application, and the applying date be that the sequence number in November 16 calendar year 2001 is No.09/988,420 U. S. application is hereby incorporated by them.Each process object in this software comprises the present invention, all is implemented as the ATL com object.

Fig. 3 represents according to the frequency response compensator of the present invention and any response generator system that are integrated in the software processes web.Note, handle the operation of web can be directly and the DSO user interactions, but provide the fundamental objects connectedness simply.In other words, Fig. 3 is illustrated in the process object connectedness in the DSO, as setting up via inner DSO software.

Among Fig. 3, in an example system architecture, illustrate according to filter part 33 of the present invention.Filter part 33 has three inputs (Input 34, Resp 35 and Corr 36) and five outputs (Output 37, Chan 38, Comp 39, Shape 40 and Noise 41).Shown input 34 is linked in the output 43 of passage 1 of acquisition system parts 42.Output pin 43 is the passage output of oscillographic digital hardware.Parts 42 constantly obtain the waveform of input 34.In this layout, the compensation waveform output of the specification that is provided to meet by digital filtering is provided in parts output 37.Output 37 is linked on the reconstructor 44, and reconstructor 44 is drawn waveform on the oscillograph screen.Response input 35 and correction input 36 provide the filtering specification that is used for determining channel response.Response input 35 is linked on the waveform input unit parts 45.In this case, the waveform input unit just reads the step response of before having obtained from same passage from disk.Proofread and correct input 36 and link on another waveform input unit parts 46, waveform input unit parts 46 are just reading the actual frequency content of step (step) response of before having obtained.The input of Resp35 and Corr 36 is made up the enough information that is used for determining the channel frequence response to filter part 33 is provided.The other parts of filtering specification are provided by the dialog box shown in Figure 15 and Figure 16 (describing after a while).Chan 38, Comp 39, Shape 40 and the output of Noise 41 filter parts provide the frequency response waveform, and these frequency response waveforms are represented the performance of system.The frequency response of the passage of determining is imported in channel response 38 outputs by Resp 35 and Corr 36.Compensation response 39 provides the compensating frequency response of digital filter, and this digital filter is designed to resist this channel frequence response.The frequency response of two filter segments of reshaper 40 and denoiser 41 outputs, these two filter segments provide the response characteristic of being stipulated by the user together in Figure 15 and dialog box shown in Figure 16.The reshaper part of digital filter is designed to mate the frequency response characteristic of regulation specially.Denoiser partly is designed to provide the rapid decay of the input waveform except that interested frequency.Chan 38, Comp 39, Shape 40 and the output of Noise 41 filter freguency response provide by decibel and can algebraic addition so that the check system performance is as shown in table 1 below.

Algebraic combination	Output (frequency response)
		Chan	Passage
Comp	The separate compensation filter segment
		Shape	Independent response generator part
Noise	Independent denoiser part
		Shape+Noise	The response of regulation
Comp+Shape+Noise	Total filter response
		Chan+Comp+Shape+Noise	The total system response of filtering
Chan+Comp	The deviation of the response of overall response and regulation (error)

Table 1

Among Fig. 3, shown four frequency responses outputs (38-41) are linked on adder part 47,48 and 49.Use reconstructor 50 and 52 to show two desired frequency response diagrams, with parts 51 and 53 come convergent-divergent they.Reconstructor 50 display channel frequency responses, and the total system frequency response of reconstructor 52 demonstrations, the i.e. summation of Chan 38, Comp 39, Shape 40 and Noise 41 outputs.

Fig. 4 illustrates the detailed view of filter part 33.Parts 33 are actually the combination of several parts.Each parts in these internal parts also all are embodied as the ATL com object of separation.Two top component (iir filter 54 and finite impulse response (FIR) (FIR) wave filter 55) are actual filter elements, and the big parts of bottom are wave filter builders 56. Wave filter input 57 and 58 is all directly linked on the input pin 59, and the DSO waveform of digital hardware is from this pin input.Equally, wave filter output 60 and 61 equal (by switches 79) are linked on the output pin 62.The setting of switch 79 or directly determine by the user, or determine by the optimization of carrying out during setting up at wave filter.

At Intel Pentium ^TMUse Intel Performance Library to realize these wave filters on the processor, Intel Performance Library contains special dual quadrant (biquad) IIR that optimizes of promising this processor and one group of dynamic link library (DLL) of FIR filter codes.Referring to Intel Signal Processing Library ReferenceMannul, No.630508-012 file, Intel company,, chapter 8 in 2000.Can utilize the Intel storehouse to realize that the fact of these wave filters only is a possible embodiment, but this for purposes of the invention not necessarily.Can adopt other any suitable realization that is equal to.(Pentium ^TMWith Intel Inside ^TMAll be the trade mark of Intel company)

The Coef output pin 65 of wave filter builder 56 is linked in Coefs input 63 and 64.Generate the coefficient of wave filter according to filter type as 62 indications of output switch.If select FIR wave filter 55, then these coefficients pressed a ₀, a ₁, a ₂, a ₃... string number send, total filter response is:

$H\left(z\right)=\underset{k}{\mathrm{\&Sigma;}}{a}_k\&CenterDot;z^{-\mathrm{<figure-callout\; id="1"\; label="k\; Equation"\; filenames="C0380952100381.png,C0380952100501.png"\; state="\{\{state\}\}">k</figure-callout>}}$ Equation 1

If select iir filter 54, then these coefficients are pressed six groups of transmissions, represent a dual quadrant part for every group.Sum is pressed a _0,0, a _1,0, a _2,0, b _0,0, b _1,0, b _2,0, a _0,1, a _1,1, a _2,1, b _0,1, b _1,1, b _2,1... send.The transport function of each part is expressed as:

$H_s\left(z\right)=\frac{a_{0,s}+a_{1,s}\&CenterDot;z^{-1}+a_{2,s}\&CenterDot;z^{-2}}{b_{0,s}+{\mathrm{<figure-callout\; id="1"\; label="b"\; filenames="C0380952100381.png,C0380952100501.png"\; state="\{\{state\}\}">b</figure-callout>}}_{1,s}\&CenterDot;z^{-1}+{\mathrm{<figure-callout\; id="2"\; label="b"\; filenames="C0380952100451.png,C0380952100481.png"\; state="\{\{state\}\}">b</figure-callout>}}_{2,s}\&CenterDot;z^{-2}}$ Equation 2

Wherein always with b _{0, s}Be arranged to 1.0

Total filter response of iir filter is:

$H\left(z\right)=\underset{s}{\mathrm{\&Pi;}}{H}_s\left(z\right)$ Equation 3

Input pin 59 is also linked in the wave filter builder input 66, is used to detect the variation of input waveform sampling rate, and this variation can need the reconstructing digital wave filter.Resp 67 and corr 68 inputs are linked in wave filter builder resp 69 and corr 70 inputs.Four frequency response output 75-78 of wave filter builder directly link on the system ensemble output 71-74.

Wave filter builder 56 needs two groups of specifications (channel frequence response and user's response specifications) to produce output parameter.The channel frequence response is used for setting up the compensated part of wave filter.User's response specifications is used for setting up any response part of wave filter.

Calculate the channel frequence response from response 69 and correction 70 input pins to wave filter builder 56.Response is the measurement response for the known input stimulus of passage.Correction is the given frequency response of reality or the frequency content of this input stimulus.As the calibration of all surveying instruments, signal source must be able to be traceable to a known standard.Therefore, known the frequency content (learning) of signal source waveform by utilizing another calibration instrument to measure, and known that this instrument to the response of this waveform (particularly, be the channel response that this treatment element is set at that DSO in its data stream), just can determine the frequency response of this passage.If H _cThe unknown frequency response of expression passage, H _sThe given frequency content of expression alignment of waveforms, H _mExpression is by the frequency content of the channel oscilloscope that does not compensate the DSO measurement, then

H _m=H _sH _cEquation 4

Thereby

$H_c=\frac{H_m}{H_s}$ Equation 5

Therefore, determine that a kind of method that channel oscilloscope responds is to adopt to have frequency content H _sKnown excitation, it is added in the input of DSO passage, obtain it with digitizer and acquisition system, measure its frequency content H _mAnd user's formula 5 is determined channel frequence response H _c

Input pin resp 69 and corr 70 are polymorphic, mean the identical interface of they expressions, but their behavior is based on input and difference.That is, each input pin can both be accepted time domain or frequency-domain waveform.Thereby this system can receive the channel frequence response specifications of following four kinds of forms:

A: the frequency sweeping of being furnished with known time domain response

Owing to seldom know the time domain response that scanning is sinusoidal wave, therefore this combination was almost never used.

B: the frequency sweeping of being furnished with the given frequency response

This combination may be modal.Find easily and supply with the sinusoidal wave exact instrument of radio frequency (RF) (for example HP8648B 2GHz signal generator of Hewlett-Packard's manufacturing).In addition, utilize RF power meter or enough accurate spectrum analyzer to measure the frequency content of the practical sinusoidal wave that flows to DSO easily.And, can utilize network analyzer to measure any frequency response characteristic that is used for carrying sinusoidal wave cable.A defective of this combination is, its needs the long time scan sine wave, and this is because interested each frequency all must flow to DSO, and must constitute measurement with the amplitude and the phase place of the signal on each Frequency point.Another defective is to be difficult to learn exactly sinusoidal wave phase place.Sometimes, the certain trigger that can utilize generator to send is exported and is overcome this difficulty.

C: the time domain waveform of being furnished with the given frequency response

This is another common combination.For the major requirement of the signal source of using this combination is that it must have enough power on interested frequency.Two common inputs are step and impulse function.Although can not easily generate fabulous step and pulse, might know the frequency content of waveform.The simplest mode is at first to come calibrated generator by obtain time domain waveform with DSO, the frequency response of using the method that discloses in the B in combination to measure passage then.Deduct the frequency content of scanning generator from frequency sweeping, add that the time-domain signal source waveform can easily calculate the frequency content of time-domain signal source waveform as measuring response.Use fast Fourier transform (FFT) or chirp Z-transform (CZT) can easily calculate the frequency response of time-domain signal source waveform.

Though the calibration of time domain generator has and the identical defective of combination described in the B, needn't carry out this continually and calibrate (just enough carrying out the calibration in time-domain signal source frequently so that remain valid).The other defective that this combination also exists is to be difficult to generate the time domain waveform that its frequency content does not change with amplitude.Because the DSO frequency response will change on its various gain margins, therefore preferably there is one can easily be used in the signal source that any possible gain is provided with.The strong point of this method is speed and measures easily.Only need input calibration time domain waveform, trigger this waveform, and enough collections are averaged so that reduce noise fully.Usually can carry out this processing with the time that is less than one second.

D: the time domain waveform of being furnished with known time domain response

This combination is seldom used.It has the identical benefit with combination C, and in a single day that is exactly to be calibrated, just can carry out the measurement of time domain waveform fast.Problem is what the actual time domain performance of signal source normally can not directly be determined.In other words, will infer it from frequency response measurement.Note, adopt DSO of the present invention, will use this combination so if in the calibration in time-domain signal source, use.

Determine to be added to the type of waveform on input resp 69 and the corr 70 by checking its waveform descriptor.Use normal linearity frequency modulation transform (CZT) that time domain waveform is transformed into frequency response.Referring to the Methods of Discrete Signal AndSystems Analysis that M.T.Jong showed, mcgraw-hill, inc, 1982, the 297-301 pages or leaves, its full content is hereby incorporated by.Use CZT to be because its allow in response, accurately to be provided with Frequency point number and no matter sampling rate.Many improved fast Fourier transform (FFT) algorithms also provide this ability, but CZT is simple, and only need radix 2 FFT and no matter input signal in counting.When the number of Frequency point in the wave filter builder is can be provided with the time, 50 points (from 0Hz to maximum compensating frequency) can be worked well.Maximum compensating frequency is that frequency that we will no longer attempt to eliminate the influence of channel frequence response thereon.Usually, this is the amplitude response of passage thereon that frequency near the noise platform.This frequency normally adopts the maximum of instrument of the present invention can reach bandwidth.

Although have fixed number point of destination (from 0Hz to maximum compensating frequency), sometimes also CZT calculated nyquist frequency limit.The performance of sometimes watching the compensated part outside interested frequency is useful.

In case the waveform transformation to the input of input resp 69 and corr 70 is become frequency response, just can determine H _sAnd H _m.Generally frequency response is expressed as amplitude (is unit with the decibel) and phase place (is unit with the degree).If desired, then use C spline interpolation to come the resampling response.Here, by deducting amplitude and phase place is calculated H _cH _cConstitute the basis of design compensation filter segment.

Fig. 5 illustrates the H that calculates _cAn example.Be used for determining H _mThe signal source waveform be the step 80 that provides by the step signal generator.This step is obtained by a DSO passage.In order to reduce noise and to increase resolution (level and vertical), with the repeatedly average step that obtains of DSO.The impulse response of desirable step is:

$\mathrm{\&delta;}\left(t\right)=\frac{d}{\mathrm{dt}}u\left(t\right)$ Equation 6

Thereby frequency content is:

$D\left(s\right)=\frac{1}{s}U\left(s\right)$ Equation 7

By the derivative of the step that obtains with a flat frequency response obtaining, and use CZT and just can easily determine step (H by a passage _s) frequency content.Fig. 5 represents the application result of equation 5.Describe frame 82 and show that the amplitude of step 80 is about 250mV, and the duration of this waveform is 20ns.The frequency response 81 that records of passage is drawn by every horizontal scale 0.5GHz, every vertical scale 1dB, as represented in the frame 83.As shown, this channel frequence response is uneven.

According to this channel response design compensation filter segment with offset response and 0dB deviation-in fact, this wave filter provides the definite inverse of channel response.At first the analog filter of this channel response is as far as possible closely simulated in design, and this wave filter is inverted so that response reciprocal to be provided, and is converted to digital filtering with bilinear transformation then.Bilinear transformation is known for those technician in the digital signal processing technique field, but still describes some details below.

Bilinear transformation is used for by direct substitution Laplce variable s analog filter being converted to digital filter.Adopt the analog filter transport function:

$H\left(s\right)=\frac{\underset{n=0}{\overset{N}{\mathrm{\&Sigma;}}}{a}_n\&CenterDot;s^n}{\underset{m=0}{\overset{M}{\mathrm{\&Sigma;}}}{b}_m\&CenterDot;s^m}$ Equation 8

Replacement below carrying out:

$s\&RightArrow;2\&CenterDot;f_s\&CenterDot;\frac{1-z^{-1}}{1+z^{-1}}$ Equation 9

And algebraic manipulation as a result equation so that part in following form:

$H\left(z\right)=\frac{\underset{n=0}{\overset{N}{\mathrm{\&Sigma;}}}{A}_n\&CenterDot;z^{-n}}{\underset{m=0}{\overset{M}{\mathrm{\&Sigma;}}}{B}_m\&CenterDot;z^{-\mathrm{<figure-callout\; id="10"\; label="m\; Equation"\; filenames="C0380952100421.png,C0380952100431.png"\; state="\{\{state\}\}">m</figure-callout>}}}$ Equation 10

By carrying out this replacement, will not exclusively carry out as the analog filter of equation 8 according to the digital filter of equation 10.This is because the nonlinear relationship between the frequency response of Analog and Digital Filters has been set up in the replacement of equation 9 expression.This nonlinear relationship is called warpage (warping).Particularly, this relation is:

$f_d=\frac{F_s}{\mathrm{\&pi;}}\&CenterDot;{\tan }^{-1}(f_a\&CenterDot;\frac{\mathrm{\&pi;}}{F_s})$ Equation 11

F wherein _dBe that frequency that numerical frequency responds estimated place, f _aBe that frequency that analog frequency responds estimated place, F _sIt is the sampling rate of digital display circuit.In other words, use this conversion, at f _aThe analog frequency response of place's estimation equals at f _dThe numerical frequency response of place's estimation.Attention:

X ≈ tan ^-1(x) for little x value equation 12

Therefore, for respect to F _sLittle f _aValue, f _a≈ f _dIn other words, for the low frequency with respect to sampling rate, the performance of digital filter is consistent with the performance of analog filter.For this reason, use the wave filter of bipolarity shift design sometimes can ignore the warpage influence.Yet in DSO, bandwidth can just in time be on the Nyquist rate.Therefore, can not ignore the warpage influence.

In order to calculate warpage, make channel frequence response warpage in advance.Fig. 6 represents pre-warpage response 201.Pre-warpage comprises the frequency scale that changes channel frequence response 200.Replace each frequency with new value and offset warpage:

$f\&RightArrow;\frac{f_s}{\mathrm{\&pi;}}\&CenterDot;\tan (\mathrm{\&pi;}\&CenterDot;\frac{f}{f_s})$ Equation 13

Notice that equation 13 trends towards infinity as f during near Nyquist rate.Even eliminating Nyquist rate, those frequencies near Nyquist still generate a large amount of pre-warpage frequencies.For this reason, the size of pre-warpage frequency is defined as a fixing multiplication coefficient (for example 50).Any pre-warpage response more than 50 times that is Nyquist rate is dropped.

Set up one and have analog filter equation 8 forms, the pre-warpage response of coupling.As what find out from pre-warpage response 201 shown in Figure 6, pre-warpage influence trends towards infinity on Nyquist.Even this means the bandwidth along with passage is exceeded, when the frequency response of passage trended towards having rapid decline, pre-warpage amplitude response still progressively flattened, near a fixed attenuation (promptly, along with response trends towards infinity, pre-warpage response is near horizontal line).The logic estimation that this means the analog filter structure is an estimation with pole and zero of similar number.For this reason, in the analog filter structure of equation 8 expressions, N=M.

By the value (filter coefficient in molecule and the denominator polynomial expression) of decision N, and to molecule and denominator coefficients a _nAnd b _mDo an initial guess and set up wave filter.Then, adjusting the mean square deviation of these coefficients between the pre-warpage channel frequence response of the amplitude response of wave filter and appointment repeatedly is minimized.Reasonable initial guess to coefficient value is very important.If not like this, L2 minimizes and can not restrain, and perhaps can replace bare minimum and converges on a local minimum.If this local minimum is away from bare minimum, the filter design that then causes can be otiose.Usually, a reasonable supposition will be any supposition that does not have overlapping pole and zero, or its frequency response approaches any supposition of channel frequence response.A Design of Filter by imagination substantially flat in the restriction that has N coefficient at wave filter designs a suitable supposition.According to Bode (Byrd) curve approximation, a single real pole or the influence of 3dB is arranged on pole location zero point.In other words, at s=-j ω _pOn a limit will be provided at f=ω _pThe decay of 3dB on/2 π.In addition, set up a flex point in the response of limit on the 3dB point.Response was smooth basically before this flex point, and every octave (octave) is fallen 6dB after this flex point.Fall the correction of an octave in any direction to this approximate 1.0dB that has an appointment.Therefore because pole and zero work is with cancellation each other, with offsetting falling of 6dB/ octave that limit sets up zero point higher on the frequency.In other words, the limit at heel zero point will be set up such response, and this response becomes flat with this limit basically, after this limit, descend with the 6dB/ octave, and or surpass the zero frequency place and flatten substantially.Thereby, if provide a pole and zero sequence, then might set up the response of a substantially flat by certain mode.This sequence can be: limit, zero point, zero point, limit, limit, zero point ...; Or zero point, limit, limit, zero point, zero point, limit ....

Approximate by checking Bode, the maximum compensating frequency of these pole and zeros and ideal flat degree should be separated an octave.Because concerning high order system, this can cause the abolishment compression to a plurality of limits under the first frequency response point in the channel frequence response, so multiplication factor-opposite with the octave separation of strictness-can be used.

Can this factor of following calculating: with terminal frequency (f _End) be defined as the last frequency in the response of pre-warpage channel frequence.With initial frequency (f _Start) be defined as a little more than 0Hz (the 8th frequency in for example pre-warpage channel frequence response).Multiplication factor (the M of pole and zero will be adapted at changing ideally between these frequencies _Space) be:

$M_{\mathrm{space}}={\left(\frac{f_{\mathrm{end}}}{f_{\mathrm{start}}}\right)}^{\frac{1}{(2\&CenterDot;N-1)}}$ Equation 14

For the octave separation of strictness, M _SpaceBe 2.0.

Consider this factor, generate a frequency array, come pole and zero is placed on these frequencies according to a kind of in two kinds of sequences of previous narration.With following formula this frequency array is described:

n∈0..2·N-1

f _n=f _Start(M _Space) ⁿ Equation 15

In case known pole and zero just can calculate molecule and the denominator polynomial expression with equation 8 forms with polynomial multiplication.

Except being the substantially flat, another characteristic that this supposition of antipodal points and null position also has is to make it to become the good initial starting point of L2 in minimizing.Because all pole and zeros (except first with last) all be adjacent along the negative real axis in the S plane, so they can become a partner easily and during wave filter match channel response as complex conjugate to leaving away.Aspect the sharp-pointed ripple in solving the channel frequence response, the complex conjugate of pole and zero is to very effective.Because plural pole and zero must occur with the conjugate pair form, therefore it is desirable to make at the very start their contiguous mutually placements on real axis.

Fig. 7 represents to have the amplitude response that the initial filtering of four pole and zeros is inferred.Fig. 7 illustrates the independent response of each limit 210 and 0. 211 and total amplitude response 212 that each contribution summation is constituted.All infer that all will contain ripple also departs from 0dB a little.Fig. 8 represents the pole and zero position of initial guess analog filter.

Must adjust now the coefficient that this initial filtering infers and respond and the error between the warpage channel response in advance so that be minimized in it.To being described below of this problem:

Suppose that pre-warpage channel frequence response contains K coordinate, each coordinate all is (ω _k, h _k) form.ω _kAnd h _kBe respectively to be the amplitude response (no unit) of the frequency and k the data point of unit with GHz.The value of obtaining a _nAnd b _mSo that mean square deviation (mse) is minimized.In other words, we minimize:

$\mathrm{mse}=\frac{1}{K}\&CenterDot;\underset{k}{\mathrm{\&Sigma;}}{\left(\right|H(j\&CenterDot;{\mathrm{\&omega;}}_k)|-h_k)}^2$ Equation 16

As filter coefficient a _nAnd b _mBe to make, when all being such coefficient of zero, just reaching (part) and minimize for the partial derivative of the mean square deviation of all these coefficients when filtered amplitude response is estimated on these coefficient values.

This be by ask gradient on it be zero that put and finish.This means that for any coefficient partial derivative all is zero:

$\frac{\&PartialD;}{{\&PartialD;a}_n}\mathrm{mse}=0$ And $\frac{\&PartialD;}{{\&PartialD;b}_m}\mathrm{mse}=0$

The estimation of these partial derivatives causes:

$\frac{\&PartialD;}{{\&PartialD;a}_n}\mathrm{mse}=\frac{2}{K}\&CenterDot;\underset{k}{\mathrm{\&Sigma;}}\left(\right|H(j\&CenterDot;{\mathrm{\&omega;}}_k)|-h_k)\&CenterDot;\frac{\&PartialD;}{{\&PartialD;a}_n}\left|H(j\&CenterDot;{\mathrm{\&omega;}}_k)\right|$ Equation 17

And

$\frac{\&PartialD;}{{\&PartialD;b}_m}\mathrm{mse}=\frac{2}{K}\&CenterDot;\underset{k}{\mathrm{\&Sigma;}}\left(\right|H(j\&CenterDot;{\mathrm{\&omega;}}_k)|-h_k)\&CenterDot;\frac{\&PartialD;}{{\&PartialD;b}_m}\left|H(j\&CenterDot;{\mathrm{\&omega;}}_k)\right|$ Equation 18

Equation 17 and 18 explanations are in order to estimate the partial derivative of mean square deviation, and we only need the analytical function of amplitude response and for the partial derivative of amplitude response.In fact, most nonlinear equation solvers all need it definitely.Amplitude response can be estimated as:

$\left|H\left(\mathrm{\&omega;}\right)\right|=\sqrt{\frac{{\mathrm{\&alpha;}\left(\mathrm{\&omega;}\right)}^2+{\mathrm{\&beta;}\left(\mathrm{\&omega;}\right)}^2}{\mathrm{\&gamma;}{\left(\mathrm{\&omega;}\right)}^2+{\mathrm{\&delta;}\left(\mathrm{\&omega;}\right)}^2}}$ Equation 19

Wherein,

$\mathrm{\&alpha;}\left(\mathrm{\&omega;}\right)=\underset{r=0}{\overset{\mathrm{floor}\left(\frac{N-1}{2}\right)}{\mathrm{\&Sigma;}}}{a}_{2\&CenterDot;r}\&CenterDot;{\mathrm{\&omega;}}^{2\&CenterDot;r}\&CenterDot;{(-1)}^{\mathrm{<figure-callout\; id="20"\; label="r\; Equation"\; filenames="C0380952100381.png,C0380952100451.png"\; state="\{\{state\}\}">r</figure-callout>}}$ Equation 20

$\mathrm{\&beta;}\left(\mathrm{\&omega;}\right)=\underset{i=0}{\overset{\mathrm{floor}\left(\frac{N}{2}\right)-1}{\mathrm{\&Sigma;}}}{a}_{2\&CenterDot;i+1}\&CenterDot;{\mathrm{\&omega;}}^{2\&CenterDot;i+1}\&CenterDot;{(-1)}^{\mathrm{<figure-callout\; id="21"\; label="i\; Equation"\; filenames="C0380952100381.png"\; state="\{\{state\}\}">i</figure-callout>}}$ Equation 21

$\mathrm{\&gamma;}\left(\mathrm{\&omega;}\right)=\underset{r=0}{\overset{\mathrm{floor}\left(\frac{M-1}{2}\right)}{\mathrm{\&Sigma;}}}{\mathrm{<figure-callout\; id="2"\; label="b"\; filenames="C0380952100451.png,C0380952100481.png"\; state="\{\{state\}\}">b</figure-callout>}}_{2\&CenterDot;r}\&CenterDot;{\mathrm{\&omega;}}^{2\&CenterDot;r}\&CenterDot;{(-1)}^{\mathrm{<figure-callout\; id="22"\; label="r\; Equation"\; filenames="C0380952100381.png"\; state="\{\{state\}\}">r</figure-callout>}}$ Equation 22

$\mathrm{\&delta;}\left(\mathrm{\&omega;}\right)=\underset{i=0}{\overset{\mathrm{floor}\left(\frac{M}{2}\right)-1}{\mathrm{\&Sigma;}}}{\mathrm{<figure-callout\; id="2"\; label="b"\; filenames="C0380952100451.png,C0380952100481.png"\; state="\{\{state\}\}">b</figure-callout>}}_{2\&CenterDot;i+1}\&CenterDot;{\mathrm{\&omega;}}^{2\&CenterDot;i+1}\&CenterDot;{(-1)}^{\mathrm{<figure-callout\; id="23"\; label="i\; Equation"\; filenames="C0380952100381.png"\; state="\{\{state\}\}">i</figure-callout>}}$ Equation 23

Partial derivative for the amplitude response of each numerator coefficients is:

$\frac{\&PartialD;}{{\&PartialD;a}_n}\left|H\left(\mathrm{\&omega;}\right)\right|=\frac{1}{2\&CenterDot;\left|H\left(\mathrm{\&omega;}\right)\right|}\&CenterDot;\frac{({\mathrm{\&gamma;}}^2+{\mathrm{\&delta;}}^2)\&CenterDot;(2\&CenterDot;\mathrm{\&alpha;}\&CenterDot;\frac{\&PartialD;}{{\&PartialD;a}_n}\mathrm{\&alpha;}+2\&CenterDot;\mathrm{\&beta;}\frac{\&PartialD;}{{\&PartialD;a}_n}\mathrm{\&beta;})-({\mathrm{\&alpha;}}^2+{\mathrm{\&beta;}}^2)\&CenterDot;(2\&CenterDot;\mathrm{\&gamma;}\&CenterDot;\frac{\&PartialD;}{{\&PartialD;a}_n}\mathrm{\&gamma;}+2\&CenterDot;\mathrm{\&delta;}\frac{\&PartialD;}{{\&PartialD;a}_n}\mathrm{\&delta;})}{{({\mathrm{\&gamma;}}^2+{\mathrm{\&delta;}}^2)}^2}$ Equation 24

Or:

Equation 25

Amplitude response to the partial derivative of each denominator coefficients is:

$\frac{\&PartialD;}{{\&PartialD;b}_m}\left|H\left(\mathrm{\&omega;}\right)\right|=\frac{1}{2\&CenterDot;\left|H\left(\mathrm{\&omega;}\right)\right|}\&CenterDot;\frac{({\mathrm{\&gamma;}}^2+{\mathrm{\&delta;}}^2)\&CenterDot;(2\&CenterDot;\mathrm{\&alpha;}\&CenterDot;\frac{\&PartialD;}{{\&PartialD;b}_m}\mathrm{\&alpha;}+2\&CenterDot;\mathrm{\&beta;}\&CenterDot;\frac{\&PartialD;}{{\&PartialD;b}_m}\mathrm{\&beta;})-({\mathrm{\&alpha;}}^2+{\mathrm{\&beta;}}^2)\&CenterDot;(2\&CenterDot;\mathrm{\&gamma;}\&CenterDot;\frac{\&PartialD;}{{\&PartialD;b}_m}\mathrm{\&gamma;}+2\&CenterDot;\mathrm{\&delta;}\&CenterDot;\frac{\&PartialD;}{{\&PartialD;b}_m}\mathrm{\&delta;})}{{({\mathrm{\&gamma;}}^2+{\mathrm{\&delta;}}^2)}^2}$ Equation 26

Or:

Equation 27

Here, known formula formula 19, equation 25 and equation 27 just can suitably be found the solution wave filter by enough any rational nonlinear equation solvers (for example genfit function among the MathCAD or Levenberg-Marquardt algorithm).

Note, when the solving equation formula, coefficient b ₀Partial derivative do not answer user's formula 27, be arranged to infinity (or maximal number) and should replace.This is because actual value a ₀And b ₀Be arbitrarily.a ₀With b ₀Ratio very important-this ratio is provided with the DC current gain of system.If one of these coefficients are unfixing, these two coefficients all can become very big or very little gradually so.By b ₀Partial derivative be set as infinity, the equation solver will be revised this parameter not obviously, and a ₀To keep unfettered so that a to be set ₀With b ₀The ratio.

Known amplitude response function and partial derivative together with the initial guess when starting filter coefficient, just can move the Levenberg-Marquardt algorithm repeatedly.Referring to the VLSI Characterization with Technology Computer-AidedDesign-PhD Thesis that Nadim Khalil is shown, Technische

Wien, 1995, its full content is hereby incorporated by.For each iteration, adjust these coefficients to reduce mean square deviation.Levenberg-Marquardt is two kinds of equilibriums between the common least squares minimization method: a kind of method of steepest descent (decent) makes along the gradient vector of mean square deviation in each iteration in this method and makes little step-length.This steepest descent method is very slow, but guarantees to converge on a local minimum.Another kind method is Newton-Gauss.Newton-Gauss (newton-Gauss) convergence is very fast, but can disperse.Levenberg-Marquardt measures the performance of himself on each iteration.Successful iteration can make it support Newton-Gauss on iteration subsequently.The iteration of failure can make it support steepest descent on iteration subsequently.The method that it is supported depends on value (λ).

Table 2

Table 2 is by the Levenberg-Marquardt algorithm iteration, and wherein g is that a coefficient vector makes:

n∈0..N

g _n=a _n Equation 28

g _n+N+1＝b _n

With mean square deviation mse ₀Be initialized as a value between initial guess filter response and the pre-warpage channel response, λ is initialized as 1000.The iteration of the method is finished when one of following condition occurring:

1. reach the mse of appointment;

2. λ reaches maximal value (for example 1e10).This expression is sometimes dispersed, but also can represent convergence;

3. reach expression system convergent minimum value (for example 1e-10);

4. on convergence point, λ may fluctuate between two or three values;

5. on convergence point, mse changes very slowly; Or

6. the iteration that exceeds maximum number.Maximal value is arranged to prevention ad infinitum carries out iteration.

In case reached local minimum, the inspection of mean square deviation just test minimizes performance.If it is not enough low, with regard to the random perturbation coefficient so that system is shaken to cause drop outside this local minimum, and so that the hope that converges on bare minimum is continued iteration.

Here, molecule and denominator multinomial coefficient are all obtained for analog filter, as describing with equation 8.This analog filter is near pre-warpage channel frequence response.Exchange the analog filter that molecule and denominator constitute the compensation channels response then.

The method of use LaGuerre makes up obtains each root of polynomial, the complex root that the method for heel Bairstow comes refinement (refine) to find with LaGuerre.Referring to the Numerical Recipes in C:the Art ofScientific Computing-second edition that William H.Press etc. is shown, the Cambridge University Press, 1992, the 369-379 pages or leaves, its full content is hereby incorporated by.Refinement comprises such hypothesis, if promptly polynomial expression is a real number, then complex root must occur with the conjugate pair form, and they are conjugate pairs.If the employing higher order polynomial, so such refinement is essential.

In case obtain root, then merge complex conjugate to and analog filter is reconstructed into:

$H\left(s\right)=\underset{\mathrm{st}}{\mathrm{\&Pi;}}\frac{a_{0,\mathrm{st}}+a_{1,\mathrm{st}}\&CenterDot;s+a_{2,\mathrm{st}}\&CenterDot;s^2}{b_{0,\mathrm{st}}+a_{1,\mathrm{st}}\&CenterDot;s+a_{2,\mathrm{st}}\&CenterDot;s^2}$ Equation 29

Wherein st is a segments of filters.Wave filter is in the form of dual quadrant section now.The number of section is more than or equal to polynomial half the smallest positive integral of initial molecule or denominator.

Can convert this wave filter to digital filter now.Use the bipolarity conversion to carry out this conversion.Each of wave filter section all is in such form:

$H\left(s\right)=\frac{\underset{n=0}{\overset{2}{\mathrm{\&Sigma;}}}{a}_n\&CenterDot;s^n}{\underset{n=0}{\overset{2}{\mathrm{\&Sigma;}}}{b}_n\&CenterDot;s^{\mathrm{<figure-callout\; id="30"\; label="n\; Equation"\; filenames="C0380952100381.png,C0380952100441.png"\; state="\{\{state\}\}">n</figure-callout>}}}$ Equation 30

For switched filter, it is alternative that we use the S shown in equation 9 to do.Do not do algebraically for band, adopt bilinearity coefficient formula to substitute and replace.Referring to the Bilinear Transform Made Easy that Peter J.Pupalaikis is shown, ICSPAT 2000 proceedings, CMP publishing company, 2000, its full content is hereby incorporated by.Each coefficient of every grade of filter section shown in the equation 30 is all converted to the digital filtering section:

$H\left(z\right)=\frac{\underset{n=0}{\overset{2}{\mathrm{\&Sigma;}}}{A}_n\&CenterDot;z^{-n}}{\underset{n=0}{\overset{2}{\mathrm{\&Sigma;}}}{B}_n\&CenterDot;z^{-\mathrm{<figure-callout\; id="31"\; label="n\; Equation"\; filenames="C0380952100381.png"\; state="\{\{state\}\}">n</figure-callout>}}}$ Equation 31

Use:

$\mathrm{BF}(i,n,N)2^i\&CenterDot;\underset{k=\max (n-N+i,0)}{\overset{\min (i,n)}{\mathrm{\&Sigma;}}}\frac{i!-(N-i)!}{k!\&CenterDot;(i-k)!\&CenterDot;(n-k)!\&CenterDot;(N-i-n+k)!}\&CenterDot;{(-1)}^{\mathrm{<figure-callout\; id="32"\; label="k\; Equation"\; filenames="C0380952100381.png"\; state="\{\{state\}\}">k</figure-callout>}}$ Equation 32

And

$A_n=\underset{i=0}{\overset{N}{\mathrm{\&Sigma;}}}{a}_i\&CenterDot;{F_s}^i\&CenterDot;\mathrm{BF}(i,n,N)$ $B_n=\underset{i=0}{\overset{N}{\mathrm{\&Sigma;}}}{b}_i\&CenterDot;{F_s}^i\&CenterDot;\mathrm{BF}(i,n,N)$ Equation 33

For the dual quadrant section, N=2 and all coefficients divided by B ₀, so that B ₀Become 1.0 and performance is constant.At this moment, calculated the compensated part of filter element.

Go up the amplitude response of this wave filter of estimation at the frequency that is used for mating the channel frequence response (point before the pre-warpage), and represent the waveform of this response and be added on the comp output pin 2 shown in Figure 4 by comp output 76 outputs of wave filter builder 56.In this way, DSO user can check the compensation filter performance.

Fig. 9 represents with the match between response that changes the consistent compensating filter of setting up and the channel frequence response.Figure 10 is illustrated in this match in the 0-2GHz scope.For this specific channel, 2GHz forces consistent maximum frequency to it, owing to be attenuated about 9dB in the response of 2GHz upper channel, so this is a rational limit.

Figure 11 represents to be designed for the amplitude response of the compensating filter that compensates this passage.This response is shown to be used for changing consistent degree.Figure 12 represents the response in the 0-2GHz scope once more.Notice that the compensating filter of Figure 12 is offset the channel frequence response.In addition, the flatness of the response that is caused is improved along with increasing consistent.Remember that consistent degree converts the filtering degree (being the dual quadrant hop count in the wave filter) of user specification to.Check Figure 12, be difficult to be clear that increase, the improvement quantity aspect compensation along with unanimity.Therefore, provide Figure 13 to illustrate, the absolute error of total bucking-out system and 0dB along with the degree that changes specified compensation filter unanimity.

Because higher consistent degree causes the more dual quadrant sections in the compensating filter, so Figure 14 is expressed as the compensating filter performance function of the progression in the wave filter.For this specific channel, be about 9dB to the 2GHz maximum error, not compensation.Average error is just above 1dB.If only compensate (promptly low consistent) with two filter sections, can even up maximum error to passage less than 0.5dB, and the average error of 0.2dB only.Utilize maximum consistent (i.e. 8 filter sections), maximum error is dropped to be lower than 0.1dB, and average error is lower than 0.04dB.Therefore, consistent degree can be used for reducing maximum error (is unit with dB) according to the amplitude of two orders of magnitude, and reduces average error according to a factor 25.

The design of any response part of wave filter is described now.Figure 15 represents a simple user interface, and this interface only is included in a control in the final response 84.This user interface allows user's nominated bandwidth 85 critically.Present DSO generally provides the selection of having only two or three fixed-bandwidths to be provided with.In addition, this user interface allows to select between four optimizations (nothing 87, impulse response 88, noiseproof feature 89 and flatness 90) of user in zone 86 is optimized in response.According to the option that is presented in this scope, may underlinedly be an additional response specifications of " especially ", it allows the user may respond from other, selects in the menu such as first order pole or the response of critical attenuation duopole.Other may response can be for special test customized.Especially, the response of stipulating by various canonical measures (for example IEEE and ANSI specified standard).Selection is used to respond the nothing 87 of optimization and turn-offs compensating filter part and response generator part.

The senior label 91 that is provided with is guided as shown in figure 16 another dialog box into.Notice that another additional control has been added to response optimization and has called support 92 times.Support the selection of noiseproof feature 93 or optimization appointment 94 to be provided.To explain this selection when being discussed below the design details of response filter.Also provide control for compensating 95.This comprises the consistent degree 96 of determining the number of dual quadrant section in the compensating filter part.And, maximum compensating frequency 97 can be set with the unanimity of assigned frequency up to expectation.Drawing in the final digital filtering realization 98 can also be provided.Show two kinds of selections, IIR 99 and FIR 100.Another possible selection is default setting (promptly automatic, this is provided with in automatic selection IIR or the FIR wave filter faster one and is used for final realization).Test shows that with regard to renewal rate iir filter always surpasses the FIR wave filter.And the pulsewidth of iir filter does not change (as FIR) with sampling rate.Therefore, for this application, iir filter is preferred wave filter, but if desired, the user also can select the FIR wave filter.Because FIR is the impulse response of blocking of IIR, therefore must regulation filtering stable (settling) amount 101 (for example 10e-6).The filtering stationary value has been determined the sampled point in the impulse response, can ignore the impulse response beyond sampled point.The stable sampling 102 of filtering is filtering stationary value according to the rules and a value calculating.Realize that for FIR this value is filtering tap (tap) number.In FIR and IIR realized, this value was that those must appear at outside the screen on the shown waveform left side so that start counting of getting ready for wave filter.

Recalling the response that is produced is made up of-Expected Response and denoiser two parts.Must comprise denoiser, be not only in order to eliminate noise, and be in order to avoid the maximum compensating frequency (f that exceeds in the compensating filter _Mc) cross to promote.This is because compensating filter is free basically outside compensating frequency.As finding out from Figure 11 and Figure 12, beyond compensating frequency, wave filter is tending towards mixed and disorderly running.Denoiser is subjected to decay that (A is set _s) and frequency configuration (f _s) control, f wherein _sBe as f _McMultiplication factor (M _Mcf) calculate.In other words, when setting up wave filter, be higher than f _McFrequency on need some decay so that compensating filter avoids mixed and disorderly running.

Discuss five kinds of feasible response optimizations according to complicated order from low to high now.Unessential situation is not optimize, and it is left to this part outside the final filtering simply, and makes compensated part useless.

Flatness optimization comprises the Butterworth Filter Design of part in response.The intention of Butterworth wave filter is to provide when influencing passband (passband) as small as possible some noises to reduce (and promote protection to crossing of compensating filter).This is designed to traditional B utterworth Filter Design, wherein in company with maximum passband decay (A _p) and minimum rejection band (stop-band) decay (A _s) together, stipulated passband and rejection band edge (f _pAnd f _s).Referring to the Digital Filter Design that T.W.Parks showed, John Wiley﹠amp; Sons company, 1987, the 159-205 pages or leaves, its full content is hereby incorporated by.Last resulting Butterworth wave filter has a rank O who calculates _ButterIf desired, this rank restrictions (clip) can be allowed O to the high-order of appointment _ButtermaxOn.If filtering is restricted to O _Buttermax, then wave filter will not meet passband and rejection band specification.In this case, the Butterworth wave filter is set to be provided at f _sAccurate decay A _sTherefore, f _pOn decay will be greater than A _pThereby, violate the flatness specification.If the filtering rank are not limited, then wave filter will meet or surpass this specification.This is because the filtering rank are chosen as the smallest positive integral that satisfies this specification.In this case, the user specifies one to depart from, and should depart from towards this in supporting specification 92 and exceed specification.If the user supports noiseproof feature 93, then the Butterworth wave filter embodies traditional design, and this design is provided at f _pOn accurate decay A _p, and provide usually and compare f _sOn A _sBetter decay.If support response to optimize 94, the Butterworth wave filter then be set to be provided at f _sAccurate decay A _sIn this case, at f _pOn decay will be less than or equal to A _p, and filtering performance will be better than flatness specification.

From user specification, derive the flatness response specifications: with f _pBe arranged to designated bandwidth frequency (f _Bw), even it is not actual bandwidth, from the specification of δ (deviation), deduct A _p, and A _sBe based on a default value of the hardware performance of special channel oscilloscope.Usually select the δ value according to typical compensating filter performance.In other words, if compensating filter can provide the unanimity of 0.1dB at most; Just may be a unnecessary constraint then less than 0.1 δ.Unless be exceeded, otherwise with f _sValue be calculated as f _McM _McfDoubly, M wherein _McfHas a default value based on special channel oscilloscope (for example 1.667).

Except with A _pBe arranged to bandwidth frequency (f _Bw) on specified attenuation (A _Bw) outside, noiseproof feature response optimization is similar to the flatness response and optimizes.Note, with A _BwDefault to 3dB, but allow to revise downwards so that guarantee bandwidth.Ignore A _sAnd f _s, and with Butterworth Design of Filter one-tenth permission (O _Buttermax) have at f _BwOn A _BwThe high-order Butterworth wave filter of decay.This just provides absolute maximum attenuation amount for given bandwidth.From user specification, derive the noiseproof feature response specifications: from bandwidth specification (f _Bw) in get f _pUnless, and be exceeded, otherwise with f _sBe calculated as f _McM _McfDoubly.

When having specified impulse response to optimize, design Besselworth wave filter is optimized response characteristic.This wave filter has the combination of Bessel and Butterworth response characteristic.The Bessel wave filter has the linear phase response characteristic and falls (roll-off) very slowly.The most important thing is that it is the low-pass filter with best pusle response characteristics.The Butterworth wave filter has the rapidest falling, and provides the response of flat passband and rejection band.Following regulation Besselworth wave filter:

1. the Bessel rank are defined as O _Bessel

2. possible maximum Butterworth rank are defined as O _Buttermax

3. it is at f _BwOn must have the A of being not more than _BwDecay;

At Bessel in frequency f _δLast response drops to A _δDB decay or be reduced to the decay of appointment following before, must depart from the Bessel response and be not more than δ; And

5. it must be at f _sOr above f _sFrequency on have A at least _sDecay.

A _BwAnd f _BwBe the bandwidth specification, δ is previous described deviation.A _δDefault value is not designated, and can be by to f _δDirectly statement heavy duty, it defaults to maximum compensating frequency f _McIn other words, default setting is suitable for the response of all tightly deferring to the Bessel response for the whole frequency range that affords redress for it.A _sAnd f _sFormerly explained.

Figure 17 represents the process flow diagram of Besselworth design process.At step 103 design simulation Bessel wave filter.Referring to the Theory and Application of Digital Signal Processing that Lawrence R.Rabiner and Bernard Gold are shown, BTL, 1975, the 228-230 pages or leaves, its full content is hereby incorporated by.Become frequency not by a specification of pre-warpage the Bessel Design of Filter.In case designed the Bessel wave filter, unless clearly stipulated f 104 _δ, otherwise just can calculate frequency f from amplitude response 105 _δ, decay reaches A on this frequency _δCalculate on Butterworth rank 106 is self-explanation, and can directly or by repetition test calculate.Note, from the decay to Butterworth requires, deducted the Bessel decay.It shall yet further be noted that and to use pre-warpage specification to determine the Butterworth rank.107 if these rank excessive, just 108 it be arranged to its maximal value.Here, 109 according to adopting the support specification, and select a kind of design (110 and 111) in two kinds of Butterworth Design of Filter with previous the same manner for flatness optimization description.In case designed this wave filter, then at 112 calculated rate f _BwOn Butterworth influence, and 113 with Bessel wave filter rescaled (by frequency) so that calculate the decay of Butterworth wave filter.Note f _δAnd f _sBe tending towards away from f _Bw, and relatively sharply falling of Butterworth wave filter generally makes it at f _BwOn influence little.This means only needs to omit inching Bessel wave filter in step 113.In addition, rescaled has been changed Bessel to be provided at f _cAnd f _δOn lower decay, but at f _sOn decay still be reduced, this ability that has endangered wave filter is to meet rejection band decay specification.Because the Bessel wave filter has slowly and falls, therefore can ignore this influence usually.A kind of method that this is compensated is when noiseproof feature is supported, will add 1 on 106 rank that calculate.Another problem is that the Butterworth wave filter may be at f _cOn big like this influence so that can not meet the bandwidth specification is arranged, promptly use high-order Butterworth.This appears at that bandwidth is specified in Nyquist rate or near Nyquist rate the time.Can be by will be at 112 A that calculate _ButterWith A _BwCompare this problem that detects.If A _ButterBigger, there is not the Bessel wave filter (because can require wave filter that gain is provided) that meets this specification so.In this case, abandon the Butterworth wave filter, system props up and uses the Bessel wave filter.In this case, the bandwidth specification is more preferably more selected than rejection band decay specification effectively.In case designed Butterworth and Bessel wave filter, just can draw Bessel filter response (pre-warpage), and make this response of analog filter match (calculating compensating filter by making compensating filter match channel response) according to similar identical mode 114.Because simulation Bessel wave filter has limit, so this analog filter has the zero point on the molecule of being added to of equal number.Use bilinear transformation that two wave filters are converted to digital filter 115.Numeral Butterworth wave filter demonstrates warpage, but has considered this warpage in its design.Because the outer simulation Bessel response of Nyquist rate is mated in this match, this Bessel wave filter fully.This method provides definite response characteristic for the Bessel filter segment.

Figure 18 represents an example of such Bessel wave filter 300.The wave filter of Figure 18 is used to have the bandwidth specification (f of 2GHz _Bw) system.Conservatively it has been defined as the decay of 2.5dB (A has been arranged on bandwidth frequency _Bw).Further it is defined as be no more than 0.5dB (δ) deviation up to such point, the second rank (O at that point _Besset=2) Bessel amplitude response 301 decay 6dB (A _δ).Owing to do not stipulate f _δ, so it be carried out the Bessel of calculating-on it response and reach-frequency of 6dB is 3.178GHz (this causes the pre-warpage specification of 7.608GHz).Found the 5th rank (O _Butter=5) Butterworth wave filter 302 can provide 20dB (A on the rejection band edge of the 3.501GHz that is calculated _s) system attenuation.The wave filter of Figure 18 meets these specifications.

Specific response-generate according to the process of summarizing among Figure 17 fully as first order pole, duopole, critical damping and other industrial standard-quilt substitutes the Bessel wave filter with specific response.In addition, Chebyshev wave filter reciprocal also is suitable for substituting the Butterworth wave filter, and this is because the pulsation (ripple) in the rejection band can be allowed rapider the ending of support certainly.

Do not consider the response filter that generates, they are converted to digital filter and keep in inside as two-stage (denoiser and reshaper).Each frequency response of output on the noise 74 of parts shown in Figure 4 and shaping 73 pins.Optimize in the situation in all responses, the Butterworth wave filter is all represented the denoiser part.In the situation that impulse response is optimized, the Bessel of Besselworth Design of Filter part is all represented shaping unit.In the specific response situation, shaping unit is represented in these responses.In the situation of flatness and noise performance optimization, do not have the wave-shaping filter part, and the frequency response of representing the unity gain on all frequencies is in 73 outputs of shaping pin.

For filtering data, this system's cascade reshaper and denoiser digital filter constitute any response and generate filter segment.Afterwards, this system's cascade compensating filter part generates filter section with any response and assigns to constitute whole compensation and response generation system.Wave filter builder 56coef output pin 65 output filter coefficients from shown in Figure 4 can use them by IIR 54 or FIR 55 wave filters here.

In a word, as follows with the interface of parts shown in Figure 4:

1. the I/O of inner filter element;

2. the specification of desired output response;

3. the specification of channel frequence response; And

4. describe the output of the pin of compensation and response, it is provided by inner filter element after wave filter builder parts have been determined wave filter.

Adopting the calibration of the system of parts shown in Figure 4 only to relate to provides the channel frequence response.Figure 19 represents to be used for the layout of the calibration of DSO 116, and DSO 116 has a probe (probe) 117 that is used for detecting tested circuit 118.Probe 117 is linked in the passage input 119 of DSO 116.Signal admission passage 120 and by the ADC digitizing, signal is processed and displayed by inner computer 121 after this.Calibration reference generator 122 is expressed as in DSO116 inside.Calibration reference generator 122 is made up of signal source 123 and calibration information 124.Reference source 122 generates such signal, and the frequency content of this signal is known.The frequency content that this is known is stored in inside as calibration data 124.This reference calibration data 124 constitutes calibration reference 122 with reference signal generator 123.Under defined terms, such as the variation of oscillograph setting, the time that changes temperature, experienced, or answer the user to ask clearly, can be by the test signal on the breaking inner input selector 125, connection connects 126 with reference to generator, control with reference to generator 123 and by the reference generator waveform of digitizing input channel 120 from then on generator obtain data and carry out calibration.Inner computer 121 deal with data collections, thus the frequency response data of measuring generated.The frequency response data of this mensuration is with being passed to the treatment element of theme of the present invention from the given frequency response 124 of calibration reference generator 122, so that determine the channel frequence response.

This calibration steps has been calibrated from the signal path of passage 120 down to switch 125, and comprises the path 126 that leads to reference to generator 123.This means path 126, and input 119 path 127 must design very carefully from switch 125 to oscillograph, perhaps must know its frequency response characteristic from switch 125 to reference generator 123.In addition, notice that probe 117 is outside calibration loop.In fact, this calibration process of being explained is only calibrated DSO at oscillograph input 119.Though might be designed to oscillographic internal path (126 and 127) high-precision, this for the probe always unfeasible.

Consider this point, many oscilloprobes have all carried the calibration information that is stored in the internal storage (EEPROM), can read this calibration information by inner computer when inserting probe.The calibration probe carries the frequency response information that can be used in the channel frequence RESPONSE CALCULATION.For example, if the frequency response of known probe, then inner computer can be added to this frequency response simply and measure in the frequency response this information being sent to before wave filter sets up parts.Resulting afterwards compensation will be taken into account the frequency response of probe.

Replacedly, the user can link probe 117 in the reference signal output 128 periodically and carry out calibration as described, except input selector switch 125 should remain on the normal operating position.The frequency response of taking from probe point 129 of passing whole passage 120 is taken into account in the calibration that is produced.Though this calibration can not full automation, it provides the highest compensativity.In addition, if this calibration is unique calibration steps that provides, so just do not need input selector switch 125 and the internal path 126 of leading to reference to generator.

In addition, calibration reference generator 122 needn't reside in the oscillograph.It can outside provide and sell as the selectable unit of DSO.When in addition, calibration data 124-is on being added to reference to generator 123-need not be arranged.These data can reside in and be used on the disk be loaded in the oscillograph.Yet, should have the certain methods of identification with reference to generator 123 and corresponding calibration data 124.According to the generator type that is adopted, perhaps inner computer is essential to the non-direct control of generator.

The wave filter builder is calculated four kinds of responses: three components of channel response and filter response.Three components of this of filter response are: compensation, reshaper and denoiser response.Utilize these response output, the arbitrary or whole algebraic combination by drawing these responses simply also offers the user to this information, just can be delivered to the user comprising the frequency response specification entirely.In this way, the user can check any frequency response behavior of being wanted.In addition, the curve map of similar Figure 13 is possible, and can provide the additional information of usefulness.In addition, can calculate various measuring (similar data shown in Figure 14) from these curve maps.

The ability that this oscillograph performance data is provided is important.For example, many canonical measures need certain surveying instrument specification (for example, a special measurement can be specified and must be used smooth oscillograph in 0.5dB outside 2GHz).The present invention not only provides the ability that satisfies this requirement, checks that final specification is to guarantee consistent ability but also provide.At last, the present invention is convenient to write down and printout oscillograph specification (as shown in figure 20) with user's measurement result, thereby the affirmation to suitable measuring condition is provided.

Though used particular term to describe the preferred embodiments of the present invention, this explanation only is used for illustration purpose, be to be understood that the spirit or scope that to make all changes and change and not break away from following claims.

Claims (35)

1. signal processing system, described signal processing system can compensate the channel response characteristic of input waveform and have the input specification able to programme of regulation Filter Design,

Wherein said input specification able to programme comprises:

The channel response characteristic is used to define the response characteristic of the passage that is used for obtaining described input waveform; And

User specification is used for the frequency response of regulation expectation and the consistent degree that responds with this desired frequency;

Wherein said signal processing system comprises:

The wave filter builder is used to described wave filter generation coefficient and exports the final properties specification, and described wave filter builder has:

The compensating filter generator is used for generating the coefficient that responds corresponding to compensation according to the inverse of channel response characteristic; And

The response filter generator is used for according to the coefficient of user specification generation corresponding to the combination of ideal response and noise reduction response; And

Wherein said wave filter is used to filter the overall response waveform that described input waveform and output device have described desired frequency response, and described wave filter comprises:

The filter coefficient high-speed cache is used to store the coefficient that is generated by described wave filter builder;

The compensating filter part, be used to use be stored in the described filter coefficient high-speed cache, filter described input waveform corresponding to the coefficient of described compensation response; And

The response filter part has response filtering stage and noise reduction level, is used to filter the compensation waveform of partly exporting from described compensating filter and exports described overall response waveform; Described response filter partly uses the coefficient that is stored in described combination in the described filter coefficient high-speed cache, that respond corresponding to described ideal response and described noise reduction to filter.

2. according to the signal processing system of claim 1, wherein described wave filter is embodied as infinite impulse response filter.

3. according to the signal processing system of claim 1, wherein described wave filter is embodied as finite impulse response filter.

4. according to the signal processing system of claim 1, wherein said channel response characteristic is based on the reference signal of calibration and by external reference signal that described passage obtained and predetermined.

5. according to the signal processing system of claim 1, wherein said user specification comprises bandwidth, response optimization, compensation unanimity and wave filter realization type.

6. according to the signal processing system of claim 5, wherein said response optimization is the impulse response optimization that utilizes the Besselworth wave filter to realize.

7. according to the signal processing system of claim 5, wherein said response optimization is the noise performance optimization of utilizing the Butterworth wave filter to realize.

8. according to the signal processing system of claim 5, wherein said response optimization is the flatness optimization that utilizes the Butterworth wave filter to realize.

9. according to the signal processing system of claim 5, wherein said wave filter realizes that type is finite impulse response (FIR) or infinite impulse response.

10. according to the signal processing system of claim 1, wherein said user specification is default to be predetermined value.

11. a Signal Processing Element that is used to filter the digital waveform of input comprises:

The wave filter builder is used for generating filter coefficient according to channel frequence response and user's response characteristic; Described channel frequence response is determined according to response input and correction input;

The infinite impulse response iir filter has the IIR coefficient input that described wave filter builder is linked in the IIR input and that is used for described input digit waveform; Described iir filter produces IIR filtering waveform according to the filter coefficient that is generated by described wave filter builder from this input digit waveform;

Finite pulse response FIR filter has the FIR coefficient input that described wave filter builder is linked in the FIR input and that is used for described input digit waveform; Described FIR wave filter produces FIR filtering waveform according to the filter coefficient that is generated by described wave filter builder from this input digit waveform; And

The outlet selector switch is used to select described IIR filtering waveform or described FIR filtering waveform for output.

12. according to the Signal Processing Element of claim 11, wherein said wave filter builder detects the variation of the sampling rate of the described input digit waveform that requires the generation filter coefficient.

13. according to the Signal Processing Element of claim 11, wherein said wave filter builder is that described FIR wave filter or described iir filter generate filter coefficient according to described outlet selector switch.

14. according to the Signal Processing Element of claim 11, wherein said wave filter builder has passage, compensation, reshaper and noise reduction output, is used to assess the performance of filtering.

15. according to the Signal Processing Element of claim 11, wherein said response input is a known input response, and described correction input is one by measurement input response that input channel obtained.

16. according to the Signal Processing Element of claim 11, wherein said user's response characteristic is used to generate the filter coefficient corresponding to any response part of this wave filter.

17. according to the Signal Processing Element of claim 11, wherein said user's response characteristic comprises bandwidth, response optimization, compensation unanimity and wave filter realization type.

18. according to the Signal Processing Element of claim 17, wherein said response optimization is the impulse response optimization that utilizes the Besselworth wave filter to realize.

19. according to the Signal Processing Element of claim 17, wherein said response optimization is the noise performance optimization of utilizing the Butterworth wave filter to realize.

20. according to the Signal Processing Element of claim 17, wherein said response optimization is the flatness optimization that utilizes the Butterworth wave filter to realize.

21. according to the Signal Processing Element of claim 17, wherein said wave filter realizes that type is FIR or IIR.

22. according to the Signal Processing Element of claim 11, wherein said user's response characteristic is default to be predetermined value.

23. a method of filtering the input digit waveform with the response characteristic of compensation acquisition channel comprises:

Use substep and response generates a wave filter according to input channel compensated part; With

Use the described compensation section of described wave filter to assign to filter described input digit waveform;

Wherein said substep comprises:

The described input channel response of pre-warpage;

Design an analog filter, this analog filter is inferred the input channel response of simulating this pre-warpage by making initial filtering, and the coefficient that the described initial filtering of iteration is inferred is to minimize mean square deviation;

Described analog filter reverses; And

Use bilinear transformation to come the analog filter of this counter-rotating of digitizing to produce the described compensation section of described wave filter.

24. method according to claim 23, comprise the steps: that further the user specification according to input generates any response part of described wave filter, wherein use described any response part of described wave filter to filter described input digit waveform, thereby produce digital waveform response characteristic, that filter with expectation.

25. according to the method for claim 24, the user specification of wherein said input comprises bandwidth, response optimization, compensation unanimity and wave filter realization type.

26. according to the method for claim 24, described any response part of wherein said wave filter comprises reshaper and denoiser.

27. according to the method for claim 24, the user specification of wherein said input is default to be predetermined value.

28., wherein described wave filter is embodied as infinite impulse response filter according to the method for claim 23.

29., wherein described wave filter is embodied as finite impulse response filter according to the method for claim 23.

30. according to the signal processing system of claim 23, the response of wherein said input channel is according to the reference signal of calibration with by external reference signal that described passage obtained and predetermined.

31. according to the method for claim 23, wherein said response optimization is the impulse response optimization that utilizes the Besselworth wave filter to realize.

32. according to the method for claim 23, wherein said response optimization is the noise performance optimization of utilizing the Butterworth wave filter to realize.

33. according to the method for claim 23, wherein said response optimization is the flatness optimization that utilizes the Butterworth wave filter to realize.

34. according to the method for claim 23, wherein said wave filter realizes that type is FIR or IIR.

35. according to the method for claim 23, wherein the coefficient that described initial filtering is inferred carries out iteration, the compensation unanimity of stipulating in less than the user specification in described input up to described mean square deviation.

Images